Most (not all) models of human preferences are rather agnostic about the process by which we form beliefs about the world. And on the occasion that scholars do think about the subject, they general posit that people change their beliefs based on the acquirement of new information.
But this can't be right -- or at least, not always. First, people have to at some point form a set of priors, which necessarily occurs mostly devoid of factually inquiry. Second, people do seem to form beliefs about certain subjects that are not typically amenable to factual support (such as a belief in God). Third, people often seem to resist updating prior beliefs even in the face of contradictory data, indicating that belief construction can be affected by considerations other than hard data. There are situations where there might be utility considerations in "selecting" between different beliefs -- for example, someone who suffers from anxiety might prefer to believe that good things will happen in the future, because they experience a utility loss from worrying about future losses. Eliminating this loss might outweigh any utility gains from forming a more accurate belief structure about the future.
In this week's WIP, Professor Anup Malani presented the findings of an experiment designed to test the possibility of forming beliefs without evidence. The trick was to use the placebo effect, which fundamentally is the statistically measurable boost one gets from taking an action that one believes will improve performance, divorced from any "objective" reasons that performance should actually be enhanced. The (rather ingenuous) model of the study, in a simplified form, is as follows:
The study involved volunteers who were subjected to a battery of mental tests before and after consuming a glucose drink. To incentivize the participants to perform well, they were given a monetary award. Some of the participants were told that the drink was associated with a placebo effect -- subjects who believed the drink would improve their performance would, in fact, see performance gains. This creates the incentive to change a belief without actually providing any evidence to warrant it: the volunteers don't have any objective data telling them that glucose drinks improve test performance, but they do have a reason to hold that belief (a desire to trigger the placebo effect and thus do better on the test).
The study found that volunteers told about the placebo effect did, in fact, self-report a greater belief in the effectiveness of glucose drinks. However, this did not translate into better performance on the test itself. The implication is that people cannot consciously choose to change their beliefs just because they have an incentive to do so.